| *** pleia2_ is now known as pleia2 | 01:04 | |
| *** yfried has joined #openstack-shade | 02:26 | |
| *** yfried has quit IRC | 03:16 | |
| *** gouthamr has quit IRC | 03:26 | |
| *** yfried has joined #openstack-shade | 04:28 | |
| *** yfried has quit IRC | 04:45 | |
| *** jamielennox is now known as jamielennox|away | 07:59 | |
| *** jamielennox|away is now known as jamielennox | 08:06 | |
| *** abregman has joined #openstack-shade | 08:32 | |
| *** yfried has joined #openstack-shade | 08:37 | |
| *** abregman is now known as abregman|afk | 10:31 | |
| *** cdent has joined #openstack-shade | 10:51 | |
| *** abregman|afk is now known as abregman | 12:00 | |
| *** purplerbot has quit IRC | 12:15 | |
| *** purplerbot has joined #openstack-shade | 12:15 | |
| *** cdent has quit IRC | 12:30 | |
| *** ioggstream has joined #openstack-shade | 13:24 | |
| *** gouthamr has joined #openstack-shade | 14:48 | |
| *** cdent has joined #openstack-shade | 15:14 | |
| *** gouthamr has quit IRC | 15:22 | |
| *** gouthamr has joined #openstack-shade | 15:29 | |
| *** gouthamr has quit IRC | 15:30 | |
| *** larainema has quit IRC | 15:46 | |
| *** yfried has quit IRC | 16:20 | |
| *** openstackgerrit has quit IRC | 16:35 | |
| *** abregman has quit IRC | 17:35 | |
| *** gouthamr has joined #openstack-shade | 18:02 | |
| *** ioggstream has quit IRC | 18:09 | |
| *** gouthamr has quit IRC | 18:35 | |
| *** greenlight_ has joined #openstack-shade | 19:32 | |
| *** gouthamr has joined #openstack-shade | 19:41 | |
| *** gouthamr has quit IRC | 19:41 | |
| *** greenlight_ has left #openstack-shade | 19:48 | |
| *** yfried has joined #openstack-shade | 20:02 | |
| *** yfried has quit IRC | 20:28 | |
| *** ianw has quit IRC | 21:07 | |
| *** ianw has joined #openstack-shade | 21:07 | |
| *** ioggstream has joined #openstack-shade | 21:20 | |
| *** gouthamr has joined #openstack-shade | 21:21 | |
| *** cdent has quit IRC | 21:36 | |
| *** jamielennox is now known as jamielennox|away | 21:42 | |
| *** gouthamr has quit IRC | 21:45 | |
| dhellmann | mordred : if I want to boot an ephemiral instance and attach a volume, is there a secret incantation? I keep getting "Invalid input for field/attribute boot_index. Value: None. None is not of type 'integer" | 22:19 |
|---|---|---|
| morgan | dhellmann: did you speak it in latin, backwards.... because......... >.> | 22:20 |
| morgan | dhellmann: sorry, let me be less snarky | 22:20 |
| morgan | i dunno off the top of my head mordred might though. | 22:20 |
| dhellmann | I'm using this ansible snippet: http://paste.openstack.org/show/598019/ | 22:20 |
| dhellmann | it worked on dreamhost, but all of their instances boot from volumes (I think) | 22:20 |
| mordred | dhellmann: the actual incantation is INSANE to do that ... but we should be hiding the crazy in shade | 22:20 |
| dhellmann | I'm using my devstack host now, and it doesn't work | 22:20 |
| morgan | mordred: oh you translate it to ancient sumerian for dhellmann then? :P | 22:21 |
| dhellmann | mordred : I'm passing stuff to shade via ansible, so... | 22:21 |
| dhellmann | this version has the volume setup too: http://paste.openstack.org/show/598020/ | 22:23 |
| mordred | dhellmann: ok. you may have found a bug | 22:23 |
| dhellmann | whee! | 22:23 |
| * mordred looking at api spec real quick | 22:23 | |
| mordred | https://github.com/openstack-infra/shade/blob/master/shade/openstackcloud.py#L5176-L5191 | 22:24 |
| mordred | is the seciton in question | 22:24 |
| mordred | (please note the wonderful format that is expected to be passed in) | 22:25 |
| dhellmann | our apis are lovely | 22:25 |
| dhellmann | what is boot_index even supposed to be? | 22:25 |
| dhellmann | well, setting boot_from_volume=yes doesn't help | 22:26 |
| mordred | http://docs.openstack.org/developer/nova/block_device_mapping.html#block-device-mapping-v2 | 22:26 |
| mordred | "Setting a negative value or None indicates that the device should not be used for booting. The simplest usage is to set it to 0 for the boot device and leave it as None for any other devices." | 22:26 |
| * mordred sighs | 22:26 | |
| mordred | so apparently that's not true anymore | 22:26 |
| dhellmann | I guess now it must be a negative value | 22:26 |
| dhellmann | let me try changing my copy of shade locally to see if -1 works | 22:26 |
| mordred | cool | 22:26 |
| mordred | I also wonder if just leaving it out of the payload works too | 22:27 |
| dhellmann | I'll try both | 22:27 |
| dhellmann | we should have some regression tests using shade in the gate | 22:28 |
| mordred | wow. from the current nova api guide: block_device_mapping_v2.boot_index body string | 22:28 |
| mordred | dhellmann: yah - turns out we have some functional tests in shade that catch a lot of regressions | 22:28 |
| dhellmann | -1 worked | 22:28 |
| mordred | \o/ | 22:28 |
| dhellmann | I'll submit that patch | 22:29 |
| mordred | thanks - I'm going to go poke the nova team to see ifthis is actually a nova bug | 22:29 |
| dhellmann | k | 22:29 |
| dhellmann | I'll try leaving it out next | 22:29 |
| dhellmann | oh, interesting, you're passing '0' elsewhere | 22:30 |
| dhellmann | rather than 0 | 22:30 |
| dhellmann | I used -1 and that worked | 22:31 |
| mordred | the docs _say_ string :) | 22:31 |
| dhellmann | mordred: leaving it out also seems to work, which version do you prefer? | 22:31 |
| dhellmann | maybe '-1' would be more backwards compatible | 22:31 |
| dhellmann | trying '-1' now | 22:32 |
| *** ioggstream has quit IRC | 22:32 | |
| dhellmann | that seems to work, too | 22:32 |
| mordred | dhellmann: my hunch is nova bug if the code worked on dreamhost, and it's in shade like that which means we _hopefully_ tested it at some point | 22:33 |
| mordred | dhellmann: but honestly, I think '-1', -1 and just leaving it out all seem like lovely solutions | 22:33 |
| mordred | whichever feels good to you | 22:34 |
| dhellmann | I guess I could test against dreamhost, too, to make sure it works there, too | 22:34 |
| clarkb | fwiw the volume attachment boot dance is one of my least favorite things in openstack right now | 22:36 |
| clarkb | there are no good consistent ways to determine where the volume is attached, and you have to do dance like things in order to partition and format and mount in the right order before things get configured to use disk | 22:37 |
| morgan | clarkb: ugh | 22:41 |
| morgan | clarkb: double ugh | 22:41 |
| morgan | clarkb: triple ugh | 22:41 |
| morgan | clarkb: unfun. | 22:41 |
| clarkb | morgan: ya its different between kvm and xen and you almost need a "pause at boot right here so that I can do all the things nthat I need to do before you attempt to do things like write logs to disk or start services or run config management" | 22:41 |
| *** jamielennox|away is now known as jamielennox | 22:42 | |
| clarkb | morgan: tldr is that i have laerned what I really want most of the time is manilla :P | 22:42 |
| clarkb | but I haven't actually been able to use it yet as I don't think any of our public clouds offer it | 22:42 |
| morgan | right | 22:42 |
| morgan | i mean... it almost sounds like something cinder should be doing vs *YAOS* (that would be yet-another-openstack-service) | 22:43 |
| morgan | unless cinder has just punted on volume attachment boot things. | 22:43 |
| mordred | dhellmann: this makes me worry that we don't have a functional test of attaching a non-boot volume at boot time | 22:43 |
| morgan | mordred: we probably dont | 22:43 |
| clarkb | morgan: I dig in a while back and cinder can format but only to swap I think | 22:43 |
| morgan | clarkb: *blink*. | 22:44 |
| mordred | morgan: well, I believe it's actualy _nova_ that does cinder volume attachment on boot | 22:44 |
| morgan | clarkb: i.. that makes me sad. | 22:44 |
| clarkb | mordred: that is when I discovered that you can use nova api as a key value store beacuse random things like image format don't actually get type checked | 22:44 |
| morgan | clarkb: i should be able to say "give me volume, size X, partition table y, format Z" | 22:44 |
| morgan | this sounds exactly like a cinder thing | 22:44 |
| mordred | clarkb: well - they've fixed the type checking now | 22:44 |
| clarkb | mordred: nice! | 22:45 |
| mordred | clarkb: starting in microversion something something | 22:45 |
| morgan | mordred: yay minor improvements | 22:45 |
| *** gouthamr has joined #openstack-shade | 22:45 | |
| morgan | (minor improvements go a long way for good UX) | 22:45 |
| mordred | they do indeed | 22:45 |
| morgan | (not being snarky there, seriously, yay) | 22:45 |
| clarkb | morgan: but ya one of our struggles is we want to boot instance and do something like format an ext4 fs on cinder volume then attach it under say /var/logs | 22:45 |
| clarkb | you can't really use config management for that beacuse by the time config management runs its too late | 22:46 |
| morgan | clarkb: the way i see it, give me a volume, size x, partition table looks like y, format z, cinder gives you a guid/uuid of the fs you can mount or of partitions/fs | 22:46 |
| morgan | whatever | 22:47 |
| morgan | a clear idea of what you'd use to mount the thing | 22:47 |
| morgan | then you have it for config management | 22:47 |
| morgan | it's a bit of a dance still, but doesn't need a *pause* go do things *unpause* boot | 22:47 |
| *** openstackgerrit has joined #openstack-shade | 22:47 | |
| openstackgerrit | Doug Hellmann proposed openstack-infra/shade master: pass -1 for boot_index of non-boot volumes https://review.openstack.org/430465 | 22:47 |
| clarkb | to make things even more exciting cloud init has tried to address this by having format and mount rules except then clouds turned it on by default making it even harder for you to mount things elsewhere... | 22:48 |
| morgan | i... ugh | 22:48 |
| morgan | nooooo | 22:48 |
| * morgan still gets the clear feeling most people who write the openstack APIs never, ever, ever, have to use them | 22:49 | |
| openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Add test of attaching a volume at boot time https://review.openstack.org/430468 | 22:52 |
| mordred | dhellmann, clarkb, morgan: ^^ let's see if that breaks without doug's patch. then if it does, we can rebase it | 22:52 |
| morgan | cool | 22:53 |
| dhellmann | mordred, morgan : https://bugs.launchpad.net/nova/+bug/1662699 if you want to watch along | 22:55 |
| openstack | Launchpad bug 1662699 in OpenStack Compute (nova) "API documentation and behavior do not match for booting with attached volumes" [Undecided,New] | 22:55 |
| mordred | dhellmann: yay! we've hopefully made openstack a little bit better today | 22:56 |
| dhellmann | one drop of water at a time | 22:58 |
| openstackgerrit | Monty Taylor proposed openstack-infra/shade master: Add test of attaching a volume at boot time https://review.openstack.org/430468 | 23:48 |
| mordred | that had an error - I want to see the actual error happen | 23:49 |
| mordred | s/an error/a pep8 error/ | 23:49 |
| mordred | clarkb: feel like popping https://review.openstack.org/#/c/430465 with a +2 or a +A? | 23:49 |
| clarkb | sure let me review | 23:50 |
| clarkb | I +2'd as you mentioned wanting to make sure tests work | 23:53 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!